Homologous gene targeting has been widely used in the past to obtain site-specific and precise genome surgery (Thomas and Capecchi, 1986, Nature, 324, 34-8; Thomas et al., 1986, Cell, 44, 419-28; Thomas and Capecchi, 1986, Cold Spring Harb Symp Quant Biol, 51 Pt 2, 1101-13; Doetschman et al., 1988, PNAS, 85, 8583-7). Homologous gene targeting relies on the homologous recombination machinery, one of the endogenous maintenance systems of the cell. Since this system has been well conserved throughout evolution, gene targeting could be used in organisms as different as bacteria, yeast, filamentous fungi, mammals, insects, and plants.
One direct application is the modulation of gene expression by modifying the regulatory sequences surrounding the gene (EP 419621; U.S. Pat. Nos. 6,528,313; 6,528,314; 5,272,071; 5,641,670). Correction of mutated genes by homologous recombination is another application (Fischer et al., 2002, Isr Med Assoc J, 4, 51-4). A deleterious mutation can often be complemented by the introduction of a wild type gene anywhere else in the genome. However, there are three major drawbacks in such an approach (random transgenesis). First, the mutated gene is still present. Certain mutation will result in a gain of function, that will not be complemented by the wild type gene, or that can at least interfere with the wild type gene. Second, gene expression often depends on very long tracts of surrounding sequences. In higher eukaryotes, these sequences can span over several hundreds of kbs, and are necessary for the precise tuning of gene expression during the cell cycle, development, or in response to physiological signals. Even though transgenic sequences involve most of the time a few kbs, there is no way they can restore a fully wild type phenotype. This problem can however be alleviated by transformation with very large sequences (BAC), but it requires additional skills. Third, random transgenesis results in insertions anywhere in the genome, with a non-nul probability of a deleterious effect: insertion in a gene will disrupt the gene or its proper regulation. Such deleterious effect have been fully illustrated recently in gene therapy trials for SCID patients (Fischer et al., precited), which resulted in cases of leukemia-like syndromes, probably as a consequence of deleterious insertions of the virus-borne transgenes.
In contrast with random transgenesis, homologous recombination allows the precise modification of a chromosomal locus: it can result in gene deletion, gene insertion, or gene replacement, depending on the targeting vector. In addition, subtle changes can be introduced in a specific locus, including the modification of coding and regulatory sequences (EP 419621; U.S. Pat. Nos. 6,528,313; 6,528,314; 5,272,071; 5,641,670; 6,063,630).
These specific advantages should make homologous gene targeting a universal tool for genome engineering, and the only safe methodology for gene therapy. However, the use of homologous recombination is limited by its poor efficiency in most cells. Although homologous gene targeting is extremely efficient in the yeast Saccharomyces cerevisiae (Paques and Haber, 1999, Microbiol Mol Biel Rev, 63, 349-404), the moss Physcomitrella patens (Schaefer and Zryd, 1997, Plant J, 11, 1195-206), certain mutant Escherichia coli strains (Murphy, 1998, J. Bacteriol, 180, 2063-71; Zhang et al., 1998, Nat Genet, 20, 123-8), and in avian cell lines such as DT40 (Buerstedde and Takeda, 1991, Cell, 67, 179-88), its efficiency remains extremely low in most cells and organisms. For example in cultured mammalian cells, such recombination events usually occur in only one in ten thousands cells which have taken up the relevant correcting or targeting DNA.
As a consequence, many approaches have been used to improve the efficiency of homologous gene targeting. Chimeraplasty (Yoon et al., 1996, PNAS, 93, 2071-6), Small Fragment Homologous Recombination (Goncz et al., 2002, Gene Ther, 9, 691-4) and Triplex Forming Oligonucleotides (Gorman and Glazer, 2001, Curr Mol Med, 1, 391-9) are as many examples. However, the most robust and efficient way to improve homologous gene targeting remains to deliver a DNA double-strand break (DSB) in the locus of interest (U.S. Pat. Nos. 5,474,896; 5,792,632; 5,866,361; 5,948,678; 5,948,678, 5,962,327; 6,395,959; 6,238,924; 5,830,729). This method improves the targeting efficiency by several orders of magnitude in mammalian cells (Donoho et al., 1998, Mol Cell Biol, 18, 4070-8; Rouet et al., 1994, Mol Cell Biol, 14, 8096-106; Choulika et al., 1995, Mol Cell Biol, 15, 1968-73; Cohen-Tannoudji et al., 1998, Mol Cell Biol, 18, 1444-8; Porteus and Baltimore, 2003, Science, 300, 763; Porteus et al., 2003, Mol Cell Biol, 23, 3558-65; Miller et al., 2003, Mol Cell Biol, 23, 3550-7) and allows gene targeting in plants (Puchta et al., 1993, Nucleic Acids Res, 21, 5034-40) and Drosophila (Bibikova et al., 2003, Science, 300, 764).
Therefore, the introduction of the double-strand break is accompanied by the introduction of a targeting segment of DNA homologous to the region surrounding the cleavage site, which results in the efficient introduction of the targeting sequences into the locus (either to repair a genetic lesion or to alter the chromosomal DNA in some specific way). Alternatively, the induction of a double-strand break at a site of interest is employed to obtain correction of a genetic lesion via a gene conversion event in which the homologous chromosomal DNA sequences from an other copy of the gene donates sequences to the sequences where the double-strand break was induced. This latter strategy leads to the correction of genetic diseases either in which one copy of a defective gene causes the disease phenotype (such as occurs in the case of dominant mutations) or in which mutations occur in both alleles of the gene, but at different locations (as is the case of compound heterozygous mutations), (WO 96/14408; WO 00/46386; U.S. Pat. No. 5,830,729; Choulika et al., precited; Donoho et al., precited; Rouet et al., precited).
However, the delivery of site-specific DSBs proved to be another challenge. It requires the use of site-specific endonucleases recognizing large sequences. Such very rare-cutting endonucleases recognizing sequences larger than 12 base pairs are called meganucleases. Ideally, one would like to use endonucleases cutting only once in the genome of interest, the cleavage being limited to the locus of interest.
In the wild, such endonucleases are essentially represented by homing endonucleases (Chevalier and Stoddard, 2001, N.A.R., 29, 3757-74). Homing endonucleases are found in fungi, algae, eubacteria and archae, and are often encoded in mobile genetic elements. Their cleavage activities initiate the spreading of these mobile elements by homologous recombination. The biology of HO (Haber, 1998, Annu Rev Genet, 32, 561-99; Haber, 1995, Bioessays, 17, 609-20), 1-Scel (Jacquier and Dujon, 1985, Cell, 41, 383-94; Fairhead and Dujon, 1993, Mol Gen Genet, 240, 170-8; Colleaux et al., 1988, PNAS, 85, 6022-6; Perrin et al., 1993, Embo J, 12, 2939-47; Plessis et al., 1992, Genetics, 130, 451-60) and 1-Tevl endonucleases (Bell-Pedersen et al., 1989, Gene, 82, 119-26; Bell-Pedersen et al., 1990, Nucleic Acids Res, 18, 3763-70; Mueller et al., 1996, Genes Dev, 10, 2158-66) are among the many paradigms for such DSB-induced recombination events.
HO and 1-SceI have been used to induce homologous gene targeting in yeast (Haber, 1995, precited; Fairhead and Dujon, 1993, precited; Plessis et al., 1992, precited; U.S. Pat. Nos. 5,792,632 and 6,238,924), in cultured mammalian cells (Donoho et al.; Rouet et al.; Choulika et al.; Cohen-Tannoudji et al., precited; U.S. Pat. Nos. 5,792,632; 5,830,729 and 6,238,924) and plants (Puchta et al., 1996, PNAS, 93, 5055-60; U.S. Pat. Nos. 5,792,632 and 6,238,924). Meganucleases have also been used to trigger various intra- and interchromosomal rearrangements based on DSB-induced homologous recombinations in bacteria (Posfai et al., 1999, N.A.R., 27, 4409-15), yeast (Paques and Haber, 1999, Microbiol Mol Biol Rev, 63, 349-404), plants (Siebert and Puchta, 2002, Plant Cell, 14, 1121-31; Chiurazzi et al., 1996, Plant Cell, 8, 2057-66; Puchta, 1999, Genetics, 152, 1173-81), insects (Rong et al., 2002, Genes Dev, 16, 1568-81) and cultured mammalian cells (Lin and Waldman, 2001, Genetics, 158, 1665-74; Liang et al., 1998, PNAS, 95, 5172-7).
Group II introns proteins can also be used as meganucleases. The biology of these proteins is much more complex than the biology of homing endonucleases encoded by group I introns and inteins (Chevalier and Stoddard, precited). The protein is involved in intron splicing, and forms a ribonucleic particle with the spliced RNA molecule. This complex displays different activities including reverse splicing (of the RNA intron in a DNA strand from the target gene), nicking (of the second DNA strand in the novel gene) and reverse transcriptase (which copies the inserted RNA into a DNA strand). The final insertion of the intron into the target gene depends on all these activities. These proteins seem to induce homologous recombination, with a DSB intermediate, when the reverse transcriptase activity is mutated (Karberg et al., 2001, Nat. Biotechnol, 19, 1162-7).
Unfortunately, this method of genome engineering by using natural meganucleases for inducing homologous recombination by a double-strand break is limited by the introduction of a recognition and cleavage site of said natural meganuclease at the position where the recombinational event is desired.
Up today, in a first approach for generating new megnucleases (artificial or man-made meganucleases), some chimeric restriction enzymes have been prepared through hybrids between a DNA-binding domain (namely a zinc finger domain) and a catalytic domain (the non-specific DNA-cleavage domain from the natural restriction enzyme Fok I), (Smith et al, 2000, N.A.R, 28, 3361-9; Smith et al., 1999, Nucleic Acids Res., 27, 274-281; Kim et al., 1996, PNAS, 93, 1156-60; Kim & Chandrasegaran, 1994, PNAS, 91, 883-7; WO 95/09233; WO 94/18313; U.S. Pat. No. 5,436,150). The resulting so-called Zinc-finger nucleases have been used to induce tandem repeat recombination in Xenopus oocytes (Bibikova et al., 2001, Mol Cell Biol, 21, 289-97), and homologous gene targeting in cultured mammalian cell lines (Porteus and Baltimore, precited) and Drosophila (Bibikova et al., precited).
Another approach consisted of embedding DNA binding and catalytic activities within a single structural unit, such as a type II restriction endonuclease. However, efforts to increase the length of recognition sequence or alter the specificity of these enzymes have resulted in the loss of catalytic activity or overall diminution of specificity due to the tight interdependence of enzyme structure, substrate recognition and catalysis (Lanio et al., 2000, Protein Eng., 13, 275-281).
Based on homing endonuclease, Chevalier et al. (2002, Molecular Cell, 10, 895-905) have also generated an artificial highly specific endonuclease by fusing domains of homing endonucleases I-Dmo I and I-Cre I. The resulting enzyme binds a long chimeric DNA target site and cleaves it precisely at a rate equivalent to its natural parents. However, this experiment leads to one endonuclease with a new specificity but it is not applicable to find an endonuclease that recognizes and cleaves any desired polynucleotide sequence.
Fusions between nucleic acids and chemical compounds are another class of artificial meganucleases, wherein DNA binding and specificity rely on an oligonucleotide and cleavage on a chemical compound tethered to the oligonucleotide. The chemical compounds can have an endogenous cleavage activity, or cleave when complexed with topoisomerases (Arimondo et al., 2001, Angew Chem Int Ed Engl, 40, 3045-3048; Arimondo and Helene, 2001, Curr Med Chem Anti-Canc Agents, 1, 219-35).
Thus, meganuclease-induced recombination appears to be an extremely powerful tool for introducing targeted modifications in genomes. In addition, the development of new meganucleases able to cleave DNA at the position where the recombinational event is desired, for example derived from Zinc-finger nucleases, or from natural homing endonucleases, would allow targeting at any given locus at will and with a reasonable efficiency.
Nevertheless, it clearly emerges from the above analysis of the prior art that the use of this technology in animals has so far been mostly limited to its applications in vitro or ex vivo in cultured cells, except in the case of Drosophila (Bibikova et al. 2003, precited), where it could be used to induce recombination in a living animal, in the germline and somatic tissues.
It would be extremely advantageous to be able to use this technology to induce recombination in a whole organism, in the somatic tissues:                This could be used for tissue-specific genome engineering in animal models or foreign sequences excision in genetically-modified organisms (once the trait depending on these foreign sequences is not useful anymore). DSBs between two tandem repeats induce very high levels of homologous recombination resulting in deletion of one repeat together with all the intervening sequences (Paques and Haber, 1999, Microbiol Mol Biol Rev, 63, 349-404), and this can easily be used for the removal of any transgene with an appropriate design.        One other major application would be the use of meganuclease-induced recombination in gene therapy. In a number of cases, an ex vivo approach could be used: precursor stem cells would be taken from the patients, healed ex vivo, and grafted back in the deficient tissue. So far, ex vivo techniques have been mostly used with blood cells in SCID and other syndromes (although random insertion was used instead of homologous recombination (Fischer et al., precited). The manipulation of stem cells makes it an attractive approach for other tissues. However, the use of meganuclease-induced recombination in toto would bypass the ex vivo steps and enlarge the range of tissues that can be treated.        
There are however two major reasons why this approach is not straightforward:                First, this would require the delivery of a meganuclease in the appropriate tissue.        Second, cells in a living organism do not necessarily behave as cultured cells or germinal cells. Cultured cells and early (and sometimes late) germ cells are dividing cells, going through G1, S, G2, and M phases. In contrast, most cells in an adult animal are differentiated cells, stuck in a G0 phase. Many results indicate and/or suggest that homologous recombination does not have the same efficiency in all phases of the cell cycle (Takata et al., 1998, Embo J, 17, 5497-508; Kadyk and Hartwell, 1992, Genetics, 132, 387-402; Gasior et al., 2001, PNAS, 98, 8411-8; Essers et al., 1997, Cell, 89, 195-204). In general, the different tissues might have distinct proficiencies for homologous gene conversions. Therefore, it is not clear whether gene targeting and meganuclease-induced genome engineering by homologous recombination could be used in whole organisms, or even for ex vivo approaches, which relies on specific cell types for which recombination proficiencies are largely unknown.        
Surprisingly, by using appropriate targeting constructs and meganuclease expression vectors, the Inventors have shown that meganucleases are indeed able to induce targeted homologous recombination ex vivo and in toto, in vertebrate somatic tissues.
Accordingly, meganucleases can be used for repairing a specific sequence, modifying a specific sequence, for attenuating or activating an endogenous gene of interest, for inactivating or deleting an endogenous gene of interest or part thereof, for introducing a mutation into a site of interest or for introducing an exogenous gene or part thereof, in vertebrate somatic tissues.
Therefore, these results establish a basis for efficient site-specific genomic manipulation in mammalian somatic tissues for experimental purposes and raise the possibility of therapeutically correcting mutations by gene targeting.